AbstractEnd-of-life care training has gaps in helping students to develop attitudes toward caring for the dying. Valid and reliable assessment tools are essential in building effective educational programmes. The Frommelt Attitude Toward the Care Of the Dying scale (FATCOD-B) is widely used to measure the level of comfort/discomfort in caring for the dying and to test the effectiveness of end-of-life care training. However, its psychometric properties have been questioned and different proposals for refinement and shortening have been put forward. The aim of this study is to get to a definitive reduction of the FATCOD-B through a valid and parsimonious synthesis of the previous attempts at scale revision. Data were gathered from a sample of 220 medical students. The item response theory approach was used in this study. Of the 14 items selected from two previous proposals for scale revision, 3 had a weak correlation with the whole scale and were deleted. The resulting 11-item version had good fit indices and withstood a more general and parsimonious specification (rating scale model). This solution was further shortened to 9 items by deleting 2 of 3 items at the same level of difficulty. The final 9-item version was invariant for gender, level of religiosity and amount of experience with dying persons, free from redundant items and able to scale and discriminate the respondents.